Title: "Open source catastrophe modeling for tropical cyclones & climate change"
Abstract: Estimating risks from extreme weather events requires considering events sufficiently rare that they are not captured in historical records. The insurance industry deals with this using “catastrophe models”, which generate catalogs of synthetic events to supplement those records. The most influential and widely used industry catastrophe models, however, are based closely on historical statistics, so that it’s difficult for them to represent climate change. They also are proprietary, limiting the extent to which their methods and results are exposed to the processes of peer review and open science generally; and they generally do not consider risks for which the industry does not have significant financial exposure. Our group has developed its own global statistical-dynamical model for tropical cyclone hazard, inspired by Kerry Emanuel's pioneering work on this subject. Our model’s synthetic storms depend on the large-scale environment, and thus can capture climate change in a plausible way using inputs from reanalyses or global dynamical models. I will show some of our first results from this model: an assessment of TC hazard in Mumbai, India, and a study of global changes in TC hazard due to climate changes predicted by CMIP5 models. The latter study raises provocative questions about epistemic uncertainty: even the sign of the changes in TC frequency is uncertain, and that propagates through to a bimodal distribution of hazard and risk. How best to interpret this is a value-laden question, unresolvable by purely scientific methods. I will argue for an interpretation in which increasing uncertainty in risk amounts, pragmatically, to an increase in risk itself. Related papers 1, 2, 3 (with emphasis on the third)